Preps

Please install and load the following packages first

install.packages("tidyverse")
install.packages("ggmap")
require(tidyverse)
require(ggmap)

Why?

ggplot

1. The basics

From base R plot function to ggplot

head(mtcars)
##                    mpg cyl disp  hp drat    wt  qsec vs am gear carb
## Mazda RX4         21.0   6  160 110 3.90 2.620 16.46  0  1    4    4
## Mazda RX4 Wag     21.0   6  160 110 3.90 2.875 17.02  0  1    4    4
## Datsun 710        22.8   4  108  93 3.85 2.320 18.61  1  1    4    1
## Hornet 4 Drive    21.4   6  258 110 3.08 3.215 19.44  1  0    3    1
## Hornet Sportabout 18.7   8  360 175 3.15 3.440 17.02  0  0    3    2
## Valiant           18.1   6  225 105 2.76 3.460 20.22  1  0    3    1
plot(mtcars$wt, mtcars$mpg)

qplot(data = mtcars, wt, mpg, geom = "point")

# Note that this would also work without 'geom'
ggplot(data = mtcars, aes(x = wt, y = mpg)) + geom_point()

Practice 1

  • Generate a dataframe dfnorm with two standard normal distributions called var1 and var2 with n = 1000 (set seed to 1)
  • Create a scatterplot with ggplot

set.seed(1)
dfnorm <- data.frame(var1 = rnorm(1000), var2 = rnorm(1000))
ggplot(dfnorm, aes(var1, var2)) + geom_point()

ggplot syntax

A good breakdown from the ggplot cheat sheet.
ggplot call

Let’s add more to the plot!

p1 <- ggplot(mtcars, aes(wt, mpg, col = factor(cyl)))
p1 <- p1 + geom_point(shape = 24, size = 2, alpha = 0.6)
p1 <- p1 + geom_smooth(method = "lm", se = F, linetype = "dashed")
p1 <- p1 + scale_color_manual(values = c(`8` = "red", `6` = "green", `4` = "blue"), 
    guide = guide_legend(title = "Cylinders"))
p1 <- p1 + ggtitle(label = "MPG vs. Weight")
p1 <- p1 + xlab(label = "Weight") + ylab(label = "Miles Per Gallon (MPG)")
p1 <- p1 + theme_bw()
p1

Practice 2

  • Use cyl (remember to convert to factor) and mpg for the boxplot
  • Add points and set the size of the points to 2 and the transparency to 0.5
  • Set color to engine configuration (vs)
  • Use theme_bw() for theme

p2 <- ggplot(mtcars, aes(factor(cyl), mpg))
p2 <- p2 + geom_boxplot()
p2 <- p2 + geom_point(aes(col = factor(vs)), size = 2, alpha = 0.5)
p2 <- p2 + theme_bw()

p2

2. ggplot components

i. Aesthetics

Aesthetics would vary depending on the geom of the plot (eg. linetype for lines), but most plots would have x, y, color, fill, alpha, group, size, label, and shape. Other than x and y, all other aesthetics could be called either inside aes() or outside depending on whether you want the visual specification to map to a specific variable or to the entire dataset.

  • Discussion: What do these plots look like?

    ggplot(data = mtcars, aes(x = wt, y = mpg, col = factor(cyl), fill = factor(cyl))) + geom_point()
ggplot(data = mtcars, aes(x = wt, y = mpg)) + geom_point(aes(col = factor(cyl), fill = factor(cyl)))
ggplot(data = mtcars, aes(x = wt, y = mpg)) + geom_point(col = "red", fill = "red")
ggplot(data = mtcars, aes(x = wt, y = mpg), col = "red", fill = "red") + geom_point()

Use different aesthetics to represent more dimensions (but don’t overkill!).

ggplot(mtcars, aes(x = wt, y = mpg, col = factor(cyl), fill = factor(cyl))) + 
    geom_point(aes(shape = factor(vs), size = 2 * (hp)/mean(hp)))

There are aesthetics that are unique to certain geoms (e.g. map_id for geom_map, ymin and ymax for geom_boxplot).

ii. Geoms

Geoms are the most essential layer of ggplot objects. There are a variety of geoms to select from: histogram, bar, point, jitter, line, text, boxplot, violin, density, density2d, polygon, path, map, etc.

Here’s an example of geom_map.

usmap <- map_data("state")
## 
## Attaching package: 'maps'
## The following object is masked from 'package:purrr':
## 
##     map
p3 <- ggplot() + geom_map(data = usmap, map = usmap, aes(x = long, y = lat, 
    map_id = region), fill = "#ffffff", color = "darkgreen", alpha = 0.5)
p3 <- p3 + theme_void()

p3

iii. Stats

Stats are used when plotting the transformation of a dataset.

Here’s how to create a density plot.

set.seed(1)
rn <- data.frame(Normal.dist = rnorm(400))

p4 <- ggplot(rn, aes(x = Normal.dist))
p4 <- p4 + geom_histogram(aes(y = stat(density)), binwidth = 0.4, col = "black", 
    fill = "#FFFFFF")
p4 <- p4 + geom_density(alpha = 0.3, fill = "#FF6666") + theme_bw()

p4

In general, you can map a transformation of a variable (instead of the variable itself) to an aesthetic by using the stat function.

This also works in 2D.

set.seed(10)
df1 <- data.frame(x = rbeta(1000, 1, 3), y = rbeta(1000, 1, 3))

p5 <- ggplot(df1, aes(x, y)) + geom_point(size = 0.3)

# Option 1
p5 + stat_density2d(aes(col = ..level..))

# Option 2
p5 + stat_density2d(aes(fill = ..density..), geom = "raster", contour = F)

# Option 3
p5 + stat_density2d(aes(fill = ..density..), geom = "tile", contour = F)

A couple things to note:
* Many of the stat functions are roughly equivalent to geom functions. For example, stat_smooth and geom_smooth can be used interchangeably. The only exception is when you want to use a non-standard geom (stat_smooth has a geom = “smooth” default option which can be modified)
* you can either use stat(density) or ..density.. for aesthetics

iv. Scales

Scales help you adjust the visualization of the aesthetics (color, fill, axes, etc.).
Back to the 2d density plot

p6 <- p5 + stat_density2d(aes(fill = ..density..), geom = "raster", contour = F)
p6 <- p6 + scale_fill_distiller(palette = "Spectral", direction = -1)
p6 <- p6 + theme_void()

p6

Practice 3

  • Create a 2d density plot with dfnorm
  • Use “tile” as geom (contour = F) and “RdYlGn” for color palette

ggplot(dfnorm, aes(var1, var2)) + stat_density2d(aes(fill = ..density..), geom = "tile", contour = F) + scale_fill_distiller(palette = "RdYlGn", direction = -1) + theme_void()

Going back to the US map.

crimes <- data.frame(region = tolower(rownames(USArrests)), USArrests)
head(crimes)
##                region Murder Assault UrbanPop Rape
## Alabama       alabama   13.2     236       58 21.2
## Alaska         alaska   10.0     263       48 44.5
## Arizona       arizona    8.1     294       80 31.0
## Arkansas     arkansas    8.8     190       50 19.5
## California california    9.0     276       91 40.6
## Colorado     colorado    7.9     204       78 38.7
p7 <- p3 + geom_map(data = crimes, aes(fill = UrbanPop, map_id = region), map = usmap)
p7 <- p7 + scale_fill_viridis_c(alpha = 0.5, begin = 0.5, end = 1, direction = -1)

p7

Practice 4

  • Use cyl (remember to convert to factor) and mpg for the boxplot
  • Add points and set the size of the points to 2 and the transparency to 0.5
  • Color the points by engine configuration (vs, blue for V and red for Straight)
  • Color the box with “lightgreen” and set transparency to 0.1
  • Use theme_bw() for theme
  • Bonus: recreate the legend

p8 <- ggplot(mtcars, aes(factor(cyl), mpg))
p8 <- p8 + geom_point(aes(col = factor(vs)), size = 2, alpha = 0.5)
p8 <- p8 + geom_boxplot(alpha = 0.1, col = "lightgreen")
p8 <- p8 + xlab(label = "Cylinder") + ylab(label = "Miles Per Gallon (MPG)")
p8 <- p8 + scale_color_manual(values = c("1" = "red", "0" = "blue"), labels = c("V", "Straight"), guide = guide_legend(title = str_wrap("Engine Configuration", width = 12)))
p8 <- p8 + theme_bw()

p8

You could also apply log and sqrt transformation to the axes (scale_x_log10()). There are also options to reverse or extend the axes, and to change x to a time series axis.

v. Position

You can also adjust positions of the aesthetics. This could mean jittering a scatter plot or creating stacked bar plots.

p9 <- ggplot(mtcars, aes(factor(cyl), mpg))
p9 <- p9 + geom_point(aes(col = factor(vs)), size = 2, alpha = 0.5, position = "jitter")
p9 <- p9 + geom_boxplot(alpha = 0.1, col = "lightgreen")
p9 <- p9 + xlab(label = "Cylinder") + ylab(label = "Miles Per Gallon (MPG)")
p9 <- p9 + scale_color_manual(values = c(`1` = "red", `0` = "blue"), labels = c("V", 
    "Straight"), guide = guide_legend(title = str_wrap("Engine Configuration", 
    width = 12)))
p9 <- p9 + theme_bw()

p9

p9 <- ggplot(mtcars, aes(factor(cyl), mpg))
p9 <- p9 + geom_point(aes(col = factor(vs)), size = 2, alpha = 0.5, position = "jitter")
p9 <- p9 + geom_boxplot(alpha = 0.1, col = "lightgreen")
p9 <- p9 + xlab(label = "Cylinder") + ylab(label = "Miles Per Gallon (MPG)")
p9 <- p9 + scale_color_manual(values = c("1" = "red", "0" = "blue"), labels = c("V", "Straight"), guide = guide_legend(title = str_wrap("Engine Configuration", width = 12)))
p9 <- p9 + theme_bw()

p9

There are a few other ways to achieve this.

geom_point(position = "jitter")
geom_point(position = position_jitter())
geom_jitter()

Practice 5

  • Recreate the following plot
  • Use the same color scheme as the previous boxplot
  • Use “dogde” position to create the side-by-side bar plot
  • Use theme_classic()

p10 <- ggplot(mtcars, aes(x = factor(cyl)))
p10 <- p10 + geom_bar(aes(y = ..count.., fill = factor(vs)), position = "dodge", alpha = 0.5)
p10 <- p10 + scale_fill_manual(values = c("1" = "red", "0" = "blue"), labels = c("V", "Straight"), guide = guide_legend(title = str_wrap("Engine Configuration", width = 12)))
p10 <- p10 + xlab(label = "Cylinders")
p10 <- p10 + theme_classic()

p10

vi. Theme

All the themes we’ve used so far are default themes. You can create your own theme from scrath or make modifications to a default theme.

Here’s an example of setting a new theme

yourtheme <- theme_bw() + theme(panel.background = element_rect(fill = "yellow"))
p10 + yourtheme

vii. Facets

Create subplots based on discrete variables.

Recall the scatter plot from mtcars:

p11 <- ggplot(mtcars, aes(wt, mpg, col = factor(cyl)))
p11 <- p11 + geom_point(aes(shape = factor(vs)), size = 2, alpha = 0.6)
p11 <- p11 + scale_color_manual(values = c(`8` = "red", `6` = "green", `4` = "blue"), 
    guide = guide_legend(title = "Cylinders"))
p11 <- p11 + ggtitle(label = "MPG vs. Weight")
p11 <- p11 + xlab(label = "Weight") + ylab(label = "Miles Per Gallon (MPG)")
p11 <- p11 + theme_bw()
p11

Now if we add a facet layer to “separate” the points by color:

p11 + facet_wrap(~factor(cyl))

Or we can apply a layered facet to “separate” both by color and by shape:

p11 + facet_grid(factor(vs) ~ factor(cyl))

viii. Other Adjustments

You can also make other adjustments to your plots, almost everything within the plot is modifiable. From changing the coordinate system (for radar plots) to adding arrows and annotations, ggplot is extremely flexible and Google is often your best friend to get it done.

To summarize, create a ggplot object follows these steps:
1. ggplot() call to create your “canvas” along with dataset and aesthetics 2. create geoms and aesthetics associated with each geom 3. adjust visual elements with scale_XX_XXX 4. create title, improve axes labels and legends 5. Other additional adjustments and apply themes

Practice 6

We are going to use the gapminder dataset and try to re-create of the plots.

install.packages("gapminder")
require(gapminder)
head(gapminder)
## # A tibble: 6 x 6
##   country     continent  year lifeExp      pop gdpPercap
##   <fct>       <fct>     <int>   <dbl>    <int>     <dbl>
## 1 Afghanistan Asia       1952    28.8  8425333      779.
## 2 Afghanistan Asia       1957    30.3  9240934      821.
## 3 Afghanistan Asia       1962    32.0 10267083      853.
## 4 Afghanistan Asia       1967    34.0 11537966      836.
## 5 Afghanistan Asia       1972    36.1 13079460      740.
## 6 Afghanistan Asia       1977    38.4 14880372      786.
  • create a plot with Life Expectancy vs. GDP per capita
  • include population as size (include show.legend = F, alpha = 0.6)
  • use subplots by continent
  • scale fill with values = country_colors
  • scale x axis to logarithm of GDP

ggplot(gapminder, aes(x = gdpPercap, y = lifeExp, size = pop)) +
  geom_point(aes(fill = country), shape = 21, show.legend = F, alpha = 0.6) +
  scale_size_area(max_size = 30) +
  scale_x_log10(limits = c(150, 120000)) + ylim(c(20, 90)) + 
  scale_fill_manual(values = country_colors, aesthetics = c("colour", "fill")) +
  facet_wrap(~ continent) +
  theme_bw()

tidyverse

The entire process of manipulating datasets and creating visualization could be streamlined with tidyverse (dplyr, tidyR, ggplot).

p12 <- gapminder %>% filter(year == 1977) %>% ggplot(aes(gdpPercap, lifeExp)) + 
    geom_point(aes(fill = continent, size = pop), shape = 21, show.legend = F, 
        alpha = 0.6) + scale_size_area(max_size = 30) + scale_x_log10(limits = c(150, 
    120000)) + ylim(c(20, 90)) + scale_fill_manual(values = continent_colors, 
    aesthetics = c("colour", "fill")) + theme_bw()

p12 <- p12 + geom_text(data = gapminder %>% filter(year == 1977, pop >= 1e+08), 
    aes(label = country), size = 5, check_overlap = T, show.legend = F)

p12

gganimate

gganimate is a great tool for animation but not super straightforward.

install.packages("gganimate")
require(gganimate)
p13 <- ggplot(gapminder, aes(x = gdpPercap, y = lifeExp, size = pop)) + geom_point(aes(fill = country), 
    shape = 21, show.legend = F, alpha = 0.6) + scale_size_area(max_size = 30) + 
    scale_x_log10(limits = c(150, 120000)) + ylim(c(20, 90)) + scale_fill_manual(values = country_colors, 
    aesthetics = c("colour", "fill")) + facet_wrap(~continent) + theme_bw() + 
    # Now comes gganimate
labs(title = "Year: {floor(frame_time)}", x = "Logarithm of GDP per capita", 
    y = "Life Expectancy") + transition_time(year) + ease_aes("linear")

animate(p13, nframes = 96)

Plotly

Plotly is even more interactive than gganimate and works well with ggplot.

install.packages("plotly")
require(plotly)
p14 <- ggplot(gapminder, aes(x = gdpPercap, y = lifeExp, size = pop, fill = continent, 
    frame = year)) + geom_point(shape = 21, show.legend = F, alpha = 0.6) + 
    scale_size_area(max_size = 20) + scale_x_log10(limits = c(150, 120000)) + 
    ylim(c(20, 90)) + scale_fill_manual(values = continent_colors) + facet_wrap(~continent) + 
    theme_bw()

plty1 <- ggplotly(p14, height = 600, width = 800) %>% animation_opts(frame = 500, 
    transition = 500, easing = "linear", redraw = FALSE)
plty1